Cross-Domain Impact Analysis made easy with System Engineering

The entire project was geared towards specific application examples from industry partners. It quickly became apparent that the changes extend into the production processes, bringing the cost/benefit question of a change even more to the fore. In impact analyses, it is important to find out as quickly as possible where the greatest risks and costs of a change are lying.

The ImPaKT reference architecture for impact analyses consists of two levels. One level is the impact analysis process, i.e. the consideration of the IT systems that carry out this process. We have developed a coordinated process model for this. The second level is the reference model of the so-called parameter space (PARS) to be analyzed with this process. It is the blueprint or template for how information is related.

The reference model is described in the MBSE modeling language SysML. It is based on the requirements and experience of the project partners as well as the long-term sustainable concepts and flexible structures of the standardized STEP data models. However, this information model contains the essential know-how about the interdisciplinary networking of requirements, functions, logic or solutions, components, documents, models, and process artifacts in a much more condensed form. It is a generic model in that it is not limited to mechanical engineering but can map all types of products consisting of mechanics, E/E, software, and services.

In contrast to existing approaches, only the information relevant for impact analyses is made available as a linked data structure in one or several cascaded parameter spaces, which ensures good performance even in the most complex environments. Much of this information is already stored digitally, but it must first be made available across departments and systems, i.e. integrated and linked in an MBSE architecture. As mentioned above, a lot of information is still stored in documents or distributed across different authoring and management systems.

One of the AI use cases that promised particularly high potential and which PROSTEP developed further is the preparation of information for complex impact analyses. We ran through this using the example of Excel lists with hundreds of requirements described in text form. These texts were analyzed using NLP (Natural Language Processing) methods to identify similarities and correlations and convert them into a MBSE-compliant requirements model. It can then be made available in the parameter space and, if required, linked to component information, function, and product structures and/or test activities and their results.

The transformation of the requirements list into a structured model did not work automatically. The AI made suggestions for possible correlations based on the determined weightings, which were then evaluated by experts, resulting in a very high-quality requirements model..

With the whole supply chain in mind, the project partners have made the solution flexible and modular. As a result, the companies do not have to model their products completely with MBSE methods but can fill the parameter space with content bit by bit, e.g. by extracting and linking the essential information from the product structure and documents in the first step. They can then use AI, for example, to build a requirements model and link it to the existing structures and nodes, and so on.